Using glanceclient one could create an empty image so sample image.size will contain None value.
>>> image = glance.images.create(name='test')
>>>
>>> image
{u'status': u'queued', u'tags': [], u'container_format': None, u'min_ram': 0, u'updated_at': u'2015-08-12T09:52:20Z', u'visibility': u'private', u'owner': u'c5deed5fa1554e7baf08557947533bc8', u'file': u'/v2/images/0a35561e-b6d3-4249-b37a-36f04b5c0826/file', u'min_disk': 0, u'virtual_size': None, u'id': u'0a35561e-b6d3-4249-b37a-36f04b5c0826',
u'size': None,
u'name': u'test', u'checksum': None, u'created_at': u'2015-08-12T09:52:20Z', u'disk_format': None, u'protected': False, u'schema': u'/v2/schemas/image'}
That is not appropriate when we are talking about metrics and will crash get_sample function on MongoDB and HBase backends because we explicitly change type of volume to float (see patch https://review.openstack.org/#/c/207567/)
I see two ways for fixing:
Add default value for 'counter_volume' field, 0 --- looks like the most suitable for None, if we are talking about metric
Or check for not None counter_volume on writing data into database to exclude None metrics from storage.
As for me second way looks better.